EUVIMEDEUVIMED entdecken
Crossref

Evolutionary Transfer Neural Architecture Search Across Spaces via Representation Learning

Boyu Hou, Liang Feng, Xuefeng Chen, Jing Tang, Kay Chen Tan, Xiaofeng Liao

IEEE Transactions on Evolutionary Computation · 2026-06 · 30 (3) · 979-993

DOI: 10.1109/tevc.2025.3565326 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Evolutionary Computation, 2026-06. Verfasst von Boyu Hou und 5 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 52 Literaturangaben. Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, Postdoctoral Fellowship Program of the China Postdoctoral Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Neural architecture search (NAS) has emerged as a crucial method for automating the design of deep learning models. Despite its potential, NAS frequently requires substantial computational and hardware resources. To mitigate these challenges, transferable NAS (TNAS) has been introduced, leveraging prior NAS results to enhance performance on new tasks. However, existing methods largely focus on knowledge transfer within identical neural search spaces, overlooking the potential for cross-domain transferability. Motivated by this gap, we explore evolutionary TNAS across heterogeneous search spaces by learning common neural representations. In particular, we introduce a novel approach that encodes both operational and topological information of neural architectures into a unified sequence using a simple tokenizer. This sequence is then processed by a variational auto-encoder, with a Transformer-based encoder to capture rich neural representations and a decoder that reconstructs the original sequence. By utilizing these latent representations, we further establish an interdomain mapping that acts as a bridge, enabling effective explicit solution transfer among diverse search spaces to enhance the evolutionary NAS process. To harness this capability, we develop an evolutionary sequential transfer optimization approach that transfers knowledge during population initialization, providing both flexibility and adaptability. To the best of our knowledge, this work serves as the first attempt in the literature exploring evolutionary TNAS across diverse spaces. Moreover, we demonstrate the utility of our method through comprehensive empirical studies using different architecture spaces, including NAS-Bench-101, NAS-Bench-201, and the DARTS search space. Our results show that the proposed method significantly enhances the adaptability and performance of NAS across varied domains.

Förderung

  • National Key Research and Development Program of China
  • National Natural Science Foundation of China
  • Postdoctoral Fellowship Program of the China Postdoctoral Science Foundation
  • Natural Science Foundation of Chongqing Municipality
  • Guangdong Provincial Project
  • Basic and Applied Basic Research Foundation of Guangdong Province
  • Guangzhou Municipal Science and Technology Bureau

Zitieren

APA

Hou, B., Feng, L., Chen, X., Tang, J., Tan, K. C., & Liao, X. (2026). Evolutionary Transfer Neural Architecture Search Across Spaces via Representation Learning. IEEE Transactions on Evolutionary Computation, 30(3), 979-993. https://doi.org/10.1109/tevc.2025.3565326

Vancouver

Hou B, Feng L, Chen X, Tang J, Tan KC, Liao X. Evolutionary Transfer Neural Architecture Search Across Spaces via Representation Learning. IEEE Transactions on Evolutionary Computation. 2026;30(3):979-993. doi:10.1109/tevc.2025.3565326

BibTeX
@article{Hou2026,
  author = {Hou, Boyu and Feng, Liang and Chen, Xuefeng and Tang, Jing and Tan, Kay Chen and Liao, Xiaofeng},
  title = {Evolutionary Transfer Neural Architecture Search Across Spaces via Representation Learning},
  journal = {IEEE Transactions on Evolutionary Computation},
  year = {2026},
  volume = {30},
  number = {3},
  pages = {979--993},
  doi = {10.1109/tevc.2025.3565326}
}
Kostenlos starten →